A new non-profit research organization called STABL has officially launched with a mission to occupy the intellectual space between two polarized extremes in the AI community. By taking both the transformative potential of advanced AI and the stubborn reality of human organizational bottlenecks seriously, STABL aims to provide evidence-based guidance for a turbulent transition. The organization’s founders argue that current discourse fails to address how societal institutions can adapt to rapid technological change without being either obliterated by it or paralyzed by it.
The Four Theses of STABL
STABL operates on four core principles that define its worldview. First, it views advanced AI as a general-purpose technology with broad societal benefits. Second, it recognizes that societal transformation is bottlenecked by human and organizational change, not just model capabilities. Third, it views existing institutions as sources of stability that nonetheless require serious reform. Fourth, it advocates for a broad portfolio of defenses, emphasizing resilience as the most neglected component of AI safety. This framework positions STABL in the top-right quadrant of a 2x2 matrix, distinct from technological determinists who ignore bottlenecks and status-quo preservers who ignore potential.
Researching Recursive Self-Improvement and Resilience
The organization is actively pursuing empirical research into recursive self-improvement (RSI) through a project called CRUX. While many evaluations focus on verifiable tasks, STABL is testing AI agents on open-ended, messy research tasks to probe the true upper bounds of capability. Early results suggest frontier agents still struggle with open-ended AI research, with expert reviewers rejecting their outputs. Beyond RSI, STABL is pushing for an actionable resilience agenda, arguing that current safety efforts overemphasize alignment and non-proliferation while underinvesting in downstream defenses like cybersecurity for critical infrastructure.
Economic Impacts and Institutional Reform
STABL also plans to study the economic impacts of AI adoption, challenging the narrative that job losses are imminent. The organization notes that while workplace adoption is widespread, aggregate unemployment rates remain stable, suggesting that bottlenecks in labor market adaptation are slowing visible transformation. They aim to shift focus from pure automation to augmentation, proposing prototypes that help workers navigate evolving AI capabilities. Their research will also examine how AI impacts specific professions, such as law and software engineering, to propose policies that keep democratic institutions in control while enabling productivity gains.
Key Takeaways
- STABL is a new non-profit founded by experts including the authors of 'AI Snake Oil,' aiming to bridge the gap between AI hype and institutional inertia.
- The organization’s 'top-right quadrant' approach treats both AI capabilities and societal bottlenecks as critical variables for understanding AI’s impact.
- Key research areas include open-ended evaluations of recursive self-improvement (CRUX) and developing concrete resilience strategies for critical infrastructure.
- An advisory board featuring Erik Brynjolfsson, Helen Toner, and Audrey Tang lends significant credibility to the organization’s interdisciplinary approach.
The Bottom Line
This is a welcome addition to a discourse that often feels like shouting matches between doomsayers and dismissals. We need more organizations that treat institutional reform as a technical problem as complex as model alignment.